{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:IEX2FZZEKAHOV745IHI3264HNC","short_pith_number":"pith:IEX2FZZE","schema_version":"1.0","canonical_sha256":"412fa2e724500eeaff9d41d1bd7b876893208306f6a8ab0716e80a7f46fcc652","source":{"kind":"arxiv","id":"2501.17030","version":1},"attestation_state":"computed","paper":{"title":"Challenges in Ensuring AI Safety in DeepSeek-R1 Models: The Shortcomings of Reinforcement Learning Strategies","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Manojkumar Parmar, Yuvaraj Govindarajulu","submitted_at":"2025-01-28T15:52:51Z","abstract_excerpt":"Large Language Models (LLMs) have achieved remarkable progress in reasoning, alignment, and task-specific performance. However, ensuring harmlessness in these systems remains a critical challenge, particularly in advanced models like DeepSeek-R1. This paper examines the limitations of Reinforcement Learning (RL) as the primary approach for reducing harmful outputs in DeepSeek-R1 and compares it with Supervised Fine-Tuning (SFT). While RL improves reasoning capabilities, it faces challenges such as reward hacking, generalization failures, language mixing, and high computational costs. We propos"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2501.17030","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-01-28T15:52:51Z","cross_cats_sorted":["cs.AI","cs.CL","cs.CR"],"title_canon_sha256":"5612d7a334bd1c3294022241af9802dab2e9418983268c085d23fe8eb5c7ec6b","abstract_canon_sha256":"2cbf31adfa0c4f827760db10fe3e369735057b8357c440433c95aeb5369b0b64"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:06:21.870443Z","signature_b64":"ME6ACS+U+TlXo9ZIfHRyF8zZzsUtPhocdjII8ZI4Gy9VPdkxadJN5mM4XPve5ThSGAZkJAWKBH9+cLAr0zUBCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"412fa2e724500eeaff9d41d1bd7b876893208306f6a8ab0716e80a7f46fcc652","last_reissued_at":"2026-07-05T10:06:21.870024Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:06:21.870024Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Challenges in Ensuring AI Safety in DeepSeek-R1 Models: The Shortcomings of Reinforcement Learning Strategies","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Manojkumar Parmar, Yuvaraj Govindarajulu","submitted_at":"2025-01-28T15:52:51Z","abstract_excerpt":"Large Language Models (LLMs) have achieved remarkable progress in reasoning, alignment, and task-specific performance. However, ensuring harmlessness in these systems remains a critical challenge, particularly in advanced models like DeepSeek-R1. This paper examines the limitations of Reinforcement Learning (RL) as the primary approach for reducing harmful outputs in DeepSeek-R1 and compares it with Supervised Fine-Tuning (SFT). While RL improves reasoning capabilities, it faces challenges such as reward hacking, generalization failures, language mixing, and high computational costs. We propos"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.17030","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2501.17030/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2501.17030","created_at":"2026-07-05T10:06:21.870080+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.17030v1","created_at":"2026-07-05T10:06:21.870080+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.17030","created_at":"2026-07-05T10:06:21.870080+00:00"},{"alias_kind":"pith_short_12","alias_value":"IEX2FZZEKAHO","created_at":"2026-07-05T10:06:21.870080+00:00"},{"alias_kind":"pith_short_16","alias_value":"IEX2FZZEKAHOV745","created_at":"2026-07-05T10:06:21.870080+00:00"},{"alias_kind":"pith_short_8","alias_value":"IEX2FZZE","created_at":"2026-07-05T10:06:21.870080+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2508.19097","citing_title":"Reasoning LLMs in the Medical Domain: A Literature Survey","ref_index":45,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IEX2FZZEKAHOV745IHI3264HNC","json":"https://pith.science/pith/IEX2FZZEKAHOV745IHI3264HNC.json","graph_json":"https://pith.science/api/pith-number/IEX2FZZEKAHOV745IHI3264HNC/graph.json","events_json":"https://pith.science/api/pith-number/IEX2FZZEKAHOV745IHI3264HNC/events.json","paper":"https://pith.science/paper/IEX2FZZE"},"agent_actions":{"view_html":"https://pith.science/pith/IEX2FZZEKAHOV745IHI3264HNC","download_json":"https://pith.science/pith/IEX2FZZEKAHOV745IHI3264HNC.json","view_paper":"https://pith.science/paper/IEX2FZZE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.17030&json=true","fetch_graph":"https://pith.science/api/pith-number/IEX2FZZEKAHOV745IHI3264HNC/graph.json","fetch_events":"https://pith.science/api/pith-number/IEX2FZZEKAHOV745IHI3264HNC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IEX2FZZEKAHOV745IHI3264HNC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IEX2FZZEKAHOV745IHI3264HNC/action/storage_attestation","attest_author":"https://pith.science/pith/IEX2FZZEKAHOV745IHI3264HNC/action/author_attestation","sign_citation":"https://pith.science/pith/IEX2FZZEKAHOV745IHI3264HNC/action/citation_signature","submit_replication":"https://pith.science/pith/IEX2FZZEKAHOV745IHI3264HNC/action/replication_record"}},"created_at":"2026-07-05T10:06:21.870080+00:00","updated_at":"2026-07-05T10:06:21.870080+00:00"}